Don't know what cluster software you are using, but in most cases the cluster software, when asked to move a service to another node, will try to bring the service down gently, then with increasing force, until it succeeds.
If a backup is running while you switch the service, then the backup server will see the client essentially shutting down the connection (it connects to the virtualized IP address, which is offlined during the cluster switch). So from the point of view of the backup server it makes at least reasonable sense to assume the client shut down the connection intentionally and to consider the backup done.
Accordingly, it will not retry when the "client" (the virtual IP) comes back online.
NetBackup does the same, BTW.